understanding human action: integrating meanings
TRANSCRIPT
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -1
Understanding human action: integrating
meanings, mechanisms, causes, and contexts.
(To appear in: Interdisciplinary research: Case studies of interdisciplinary understandings of complex problems; Repko, A. F., Newell, W. H., Szostak, R. (Eds), Sage Publishers (2011). Final draft (April 2010), do not circulate without permission.)
Machiel Keestra (University of Amsterdam)
Introduction
Humans are capable of understanding an incredible variety of actions
performed by other humans. Even though these range from primary biological actions
like eating and fleeing, to acts in parliament or in poetry, humans generally can make
sense of each other’s actions. Understanding other people’s action, called “action
understanding,” can transcend differences in race, gender, culture, age, and social and
historical circumstances. Action understanding is the cognitive capability of making
sense of another person’s action by integrating perceptual information about the
behavior with knowledge about the immediate and socio-cultural contexts of the
action and with one’s own experience.
Since it is necessary to integrate multiple sources of information, it is not
surprising that failures to understand a person’s behavior are also common. Well
known is the case of an autistic professor who compares herself to an “anthropologist
from Mars.” Incapable of spontaneously understanding why someone cries, she has
learned rules that help her to infer that people who rub their eyes while tears are
running down their cheeks are weeping and probably feel unhappy (Sacks, 1995). By
contrast, normal individuals automatically allow stereotypes, prejudices, self-interests
and the like to influence their understanding a person’s behavior (Bargh & Chartrand,
1999). More generally still, humans can easily misunderstand unfamiliar symbolic
Interdisciplinary work also helps to fight individual limitations. This chapter has benefitted greatly both in wording and thought from the meticulous help of the editors of the book, Allen Repko, Bill Newell and Rick Szostak. Needless to say that remaining errors still reflect my individual limitations.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -2
actions or rituals if they rely too much on their own socio-cultural expertise
(Gadamer, 2004). Given the importance of action understanding in every domain of
human life and society, and in light of the complexities that surround it, a
comprehensive scientific understanding of this capacity is needed. Apart from
satisfying intellectual curiosity, such insight would serve to improve our action
understanding and mitigate several forms of misunderstanding. Indeed, in studying
action understanding, “we as scientists are engaged in the very process that is central
to our concerns” (Gergen & Semin, 1990, p. 1).
Scholars are increasingly dissatisfied with mono-disciplinary approaches to
understanding human action. Such one-sidedness can rest upon various motives. For
example, “hermeneutic interpretations” of action understanding tend to emphasize
historical and cultural influences while overlooking that ultimately such influences
depend upon individual cognitive processes.1 This has led to a critique of the
assumption that humans are born as a “blank slate” and that culture is solely
responsible for all cognitive contents. However, such critique easily slides into an
overemphasis on the biology of human nature and a denial of socio-cultural
influences on cognition (Pinker, 2003).
Fortunately, recent interdisciplinary endeavors have shown that an
interdisciplinary approach is preferable when investigating complex functions like
action understanding. Such research often involves developing a new “interdiscipline”
such as cultural psychology (Bruner, 1990) or combining insights from the social
sciences and psychology (Shore, 1996; Sperber, 1996). Evidence shows that
throughout human evolution there have been mutual influences between biological
and cognitive processes that shape human capacities and socio-cultural influences on
those processes (Bogdan, 2003; Donald, 1991; Tomasello, 1999). In addition to these
interdisciplinary investigations, computational sciences and artificial intelligence
research are developing computer models of human understanding that allow new
types of experiments and simulations (Churchland, 1995). Such insights underscore
the necessity and fruitfulness of disciplinary boundary crossing and require that
various disciplinary methods, concepts and theories be combined in innovative ways.
1 There is little room for evidence from the natural and social sciences in the hermeneutics as proposed in the influential (Gadamer, [1960] 2004). A theory of interpretation that allows scientific explanation a role in interpretation is proposed in (Ricoeur, [1986] 2008). In the social sciences an influential approach considers human functions as stemming from actor-network interactions, without specific interest in biological and psychological conditions (Bourdieu, [1980] 1990).
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -3
Presently, there is a need for a theoretical framework that is capable of
explaining a phenomenon as complex as human action. Such a framework requires
integrating insights from multiple disciplines. The purpose of this chapter is to
propose a “mechanism-based explanation”2 of action understanding which will
provide a theoretical framework for integrating various and often conflicting
disciplinary insights. Proposing an integrative theoretical frame is a common practice
in the sciences. Such a frame allows scientists to explain many facts that have been
observed, while predicting others. In the life and cognitive sciences, however, a
specific integrative device is often applied, namely “mechanism-based explanation.”3
As used here, “mechanism” means “an organized system of component parts and
component operations. The mechanism’s components and their organization produce
its behavior, thereby instantiating a phenomenon” (Bechtel, 2005).
In explaining action understanding, scientists assume that there is a complex
cognitive “mechanism” that is responsible for this phenomenon. Such a cognitive
“mechanism” can ‘produce’ action understanding as it processes multiple sources of
external information. Moreover, external influences can modulate or affect the
mechanism itself as is the case with socio-cultural information. For instance,
neuroimaging experiments show that the differences in family relations are correlated
with differences between the neural processes of Western and Chinese students when
they think about their mothers. In Western students, self-related thought activates
different processes than mother-related thought, while in Chinese students the two
processes are rather similar (Han & Northoff, 2008). Since action understanding
involves many more different sources of information, a mechanism-based explanatory
approach should be prepared for integrating insights from various disciplines.
2 Such an explanation is also called “mechanistic” in the philosophy of science literature (Machamer, Darden, & Craver, 2000). It is important to realize that in an explanatory context a mechanism is an epistemic device and plays a role in the organization of knowledge. If our knowledge of a phenomenon changes, the mechanism needs adjustments accordingly. 3 Since it proved to be extremely rare to demonstrate analogues of Newton’s mechanical laws for biological or cognitive systems, an alternative scientific device is considered more apt for these fields (Bechtel & Richardson, 1993; Machamer, Darden, & Craver, 2000). Meanwhile, social scientists are discussing the fruitfulness of a mechanism-based approach as well – see (Hedstrom & Swedberg, 1996), for example.
Mechanism-based Explanation in Brief
A simple and familiar example of a mechanism is a clock with components
like gears and shafts, and operations like turning and oscillating. If made properly and
provided with external inputs such as energy and correct initial settings, the clock will
establish time accurately. However, we cannot identify the mechanism that makes the
clock work just by observing its external pattern of behavior. To do that requires
going inside the clock and investigating its various components and operations.
Complex mechanisms may be analyzed at various levels. The human body, for
example, is a far more complex mechanism than a clock and must be analyzed at
various levels—anatomical, physiological, or biochemical—to be fully understood.
Each of these levels refers to the hierarchy of the body’s organization, not to the
physical size of the parts that exist at each level. Given the many and often non-linear
interactions between, for example, chemical substances, organ functions, and socio-
cultural meanings, that together can produce specific hallucinations, biological
phenomena are very complex. Compared to the human body, a mechanical clock is
not complex: underneath the observable level of shafts and gears is the unobservable
level of molecules. Note that molecular differences between clocks made from steel
or from silver do not affect the way they establish time whereas changing molecules
in blood will affect human bodily functions. Biological and cognitive mechanisms are
also far more complex than engineered mechanisms because of the non-linearity of
many intrinsic activities and their responsiveness to environmental factors, including
the meanings of socio-cultural settings and symbols.
Two strategies are used to develop a mechanism-based explanation of a
phenomenon: decomposition and localization (Bechtel & Richardson, 1993).
Decomposition means that we first analyze a given phenomenon—be it establishing
time or action understanding—into components or smaller tasks that in concert are
responsible for it. Localization means that we try to localize these components of the
phenomenon somewhere in the object or organism that displays it. In easy cases we
can localize the components of our phenomenon, like pointing the hours or the
minutes, in separate component parts and activities of the clock. However, these parts
and activities are not completely separable since they rely on the same energy source
and initial settings and share many other parts and activities. Typically, therefore, our
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -4
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -5
research leads to increasing specification and revision of the decomposition and
localization of the phenomenon with which we started. For readers who may be
unfamiliar with this approach, some clarifications are in order.
The first is that a mechanism-based explanation is not a complete description
of a clock, an animal or a brain. Rather, it is an explanation of a specific phenomenon,
event or behavior that is produced by the organized interaction of components and
operations. A mechanism-based explanation of action understanding as performed by
the brain will therefore contribute only in a limited sense to explanations of other
functions of the brain. Since a mechanism-based explanation could be given for each
function and for its components, a complete description would consist of an
unmanageable multitude of mechanisms, many of which would overlap and modulate
each other. Fortunately, explaining a specific phenomenon does not require this.
The second clarification is that a phenomenon may appear singular and
opaque, but if we are to give a (mechanism-based) explanation of it we must establish
that it is produced by different components and operations. Cognitive operations are
often called computations. These can be very simple, like addition, or more complex,
like face recognition. Figure 1.1 (from (Craver, 2007)) shows a schema of a
phenomenon, the activity of SΨ-ing.4 It also shows components X 1-4 that by
interacting in response to an external input produce the phenomenon (shown by the
four ovals below the solid surface). The arrows indicate the interactions that connect
the components, consisting mostly of simple activation or inhibition signals. These
interactions often include feed-back and feed-forward interactions between the
components and their operations. However, note that merely observing the
phenomenon (from the top down) does not reveal the complex mechanism and its
operation that produce the phenomenon. What appears on the surface to be a single
phenomenon is, in fact, a “distributed network” of smaller actions. Note also that the
phenomenon receives input (left-side arrow) and produces output, as do the
components and operations at a lower level. Explaining action understanding means
that we examine the cognitive processes that the human brain performs at various
levels and that together form the person’s capacity to understand the action or
behavior of another person.
4 If the phenomenon is complex, it is useful to de-compose such a phenomenon in sub-tasks, as we will do with action understanding. At the end of the chapter you can find a figure of this phenomenon.
Fig. 1. A phenomenon and its mechanism (from Craver, 2007, p. 7).
The third clarification is that such a mechanism can be a multi-level system
and can accordingly be examined at different levels. Obviously, we can study action
understanding while remaining at the personal level, where we observe which types of
action a person can and cannot understand, and examine the conditions that influence
his action understanding. Going into the brain to a first sub-personal level, we can
investigate which neural networks must cooperate to perform this function
appropriately. Going down to a second level, we can investigate isolated components
and activities in a particular neural area: its neurons, their interactions, and their
connections to neurons in other locations. If we need to be even more specific about
these neuronal activities, we can focus at a third level and describe the neurochemical
activities by which neurons pass on information to each other. Going in the other
direction, we can also climb one level upwards and consider the person as a
component: that is, consider him or her as a member of various social groups. At that
level, we are interested in the interactions between individuals and how they influence
each other’s action understanding, for example. For reasons discussed later, we don’t
need always to descend or ascend many levels when we explain a phenomenon like
action understanding. The study of neurochemical interactions at the third level may
still be relevant, but it is implausible that going as deep as the quantum mechanic
level of the human brain yields useful insights into action understanding.
The fourth clarification, and one that is particularly relevant to all cognitive
processes, is that mechanisms do not operate in complete isolation but are responsive
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -6
to various factors, including contextual factors. Organisms are open to external
information via their senses, but not always equally so because their motivation state,
attention and other internal processes influence this openness. Thus, the mechanism
governing action understanding will be influenced by a host of contextual variables.
The fifth is that organismic mechanisms are much more flexible systems than
other mechanisms. In organisms, we may observe that over time and due to learning
or development or to injuries, mechanisms responsible for a particular behavior have
changed or have been adapted—something a clock cannot do. Strikingly, an organism
may even develop different ways to produce the same behavior or phenomenon.
Automatization of a skill leads, for instance, to diminished involvement of conscious
control of movements, making it possible to perform other cognitive tasks
simultaneously. This can be made visible with the help of brain imaging techniques,
where experts and novices in a particular skill display strikingly different brain
activation patterns when performing similar tasks (Poldrack, et al., 2005).
A mechanism-based explanatory approach then, is particularly useful to
interdisciplinarians because it allows them achieve a more comprehensive
understanding of phenomena such as action understanding. In applying this approach,
interdisciplinarians can connect the mono-disciplinary insights in specific components
and operations at multiple levels and their intricate interactions that contribute to
human action understanding. How this works is the subject of the next section.
Part A: Drawing on disciplinary insights
Generally speaking, scientific efforts allow us ways to represent reality and
intervene in it (Hacking, 1983). Scientists represent reality by using mathematical
formulas, graphs, charts, mechanism-based and verbal explanations and the like.
Scientists intervene to test the adequacy of their representation or the predictions they
derive from it. Depending on the discipline and the representations, such interventions
range from digging for fossils in geological strata, sending particles through a
cyclotron, subjecting people to experimental conditions and adjusting variables in
computational programs used for simulations of phenomena. Choosing an adequate
type of representation of our insight into a phenomenon is an important matter, as is
the choice of an appropriate intervention to test it.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -7
Engaging in interdisciplinary research is an even more demanding process. A
mechanism-based explanation allows us to assign disciplinary insights more or less to
particular levels: neuroscientists will focus on the neuronal and neural level,
psychologists at the higher level of action understanding and its components, while
sociologists will focus on the interactions between individual persons which influence
the properties of this understanding. The challenge for interdisciplinary integration is
to demonstrate how the components and activities that occur at different levels
interact with those at other levels. However, first we must decide which disciplines
are relevant for explaining action understanding. Relevance is thus a key term for this
first part of the interdisciplinary process, if only to keep it manageable.
1. Define the problem or state the focus question: decomposition
of action understanding
Action understanding is the subject of many disciplines. This is partly the
result of it being such a general and wide-ranging phenomenon with many different
properties. While acknowledging its variability, it is useful to formulate a general
definition. In what follows, we will consider action understanding as the result of
cognitive processes that an individual -- partly unconsciously -- performs when
making sense of another person’s actions. In doing so, one person has not only to
recognize the other person is acting, but also to include various sources of information
to interpret that action. As a result, the action can be understood, and perhaps
responded to appropriately .
By putting cognitive processes at the heart of the definition we will focus on
the cognitive information processing that goes on in the brain, for which we will
establish a mechanism-based explanation. This decision is in line with recent
developments in both the cognitive and social sciences. Indeed, we may even speak of
a “cognitive turn” in many disciplines. For instance, anthropologist Bradd Shore
argues for a “cognitive view of culture” (Shore, 1996, p. 39), concurring with his
fellow social scientist Sperber who argues for combining psychology with the study
of culture because of our “psychological susceptibility to culture” (Sperber, 1996, p.
57, italics in original). In accordance with that susceptibility, another author analyzes
the human mind as a ‘neurohermeneutic system’, for which “‘[i]nterpretation’ is the
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -8
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -9
operation of neurons to represent, and act upon, reality” (Reyna, 2002, p. 112).
Finally and more extreme is the argument that even “philosophical theories are
largely the product of the hidden hand of the cognitive unconscious” (Lakoff &
Johnson, 1999, p. 14). Note that putting cognition at the heart of these approaches
does not imply that there is no room left for external and socio-cultural influences on
action understanding. Nor does it imply that the culture-specific meaning of words
and symbols don’t matter. They do, and the study of cultural determination of
meaning is highly relevant. However, for such socio-cultural aspects to have an
influence on action understanding, they must exert this influence by affecting
cognitive processes that go on in the brains of individual persons.
Having defined action understanding and put cognition at its center, we now
have to take an important step. I mentioned in the previous section that we must apply
two heuristics, decomposition and localization. Ergo, we should first try to
decompose action understanding into smaller (sub-)phenomena that can be studied
more or less separately and subsequently integrate the results. Associated with that
decomposition is our localization effort that involves finding responsible cognitive or
brain processes that do the work. In fact, we have already localized action
understanding very broadly in the individual’s cognitive processes.
After having defined action understanding as a cognitive process, we can now
decompose it further by classifying it according to its contents.5 We can follow the
lead of hermeneutic philosopher Paul Ricoeur, who has devoted much of his work to
the theory and method of interpretation of human narrative and human action. Ricoeur
pointed out that we can approach action with a set of interrelated questions, focusing
on respectively: who, what, why, how, where, when?6 His analysis prioritizes three of
those questions, which offer three different perspectives on action: ‘what’ is the
action, ‘why’ is the action being done, and ‘who’ is the agent? Since ‘who’ the agent
is refers to the agent’s identity, social roles, continuous maturation and the like, this
information will generally not be captured by the other perspectives. For that reason,
5 Even though it is very well known that classifications and taxonomies often need corrections, revisions, or additions, they are extremely helpful in delineating otherwise overwhelming domains (Dupre, 2001).5 In the case of human capacities like action understanding, our preliminary classification inevitably starts from the use of our common vocabulary. This does not imply that such words that we commonly apply hold up to scientific scrutiny. Indeed, although we may expect that concepts in the domain of human experience are more robust than concepts like ‘ether’ or ‘vital force’, we may need to revise the former, too (Machiel Keestra & Cowley, 2009). 6 This set of questions has also been proposed in the context of the classification of scientific theories in (Szostak, 2004).
Ricoeur deplores “[t]he use of ‘why?’ in the explanation of action […] became the
arbiter of the description of what counts as action” (Ricoeur, 1992, p. 61). It will
become clear that to identify the relevant contexts, we need to know more about the
agent ‘who’ performed the action. Ricoeur’s emphasis upon the ‘who’ of action does
not imply that contexts do not matter, but it is a consequence of the fact that contexts
have to make themselves felt via an individual’s cognitive processing.
So, following the lead offered by a hermeneutic analysis of understanding
actions, we can decompose action understanding into three different, but probably
interrelated, component tasks: understanding what an action is, understanding why an
action has been performed, and a more thoroughgoing understanding of the ‘who’
behind it. Evidence does indeed confirm the possibility of disentangling these three
components of action understanding that have to do with, respectively, action
recognition, intention understanding, and narrative understanding.
2 Justify using an interdisciplinary approach: action
understanding as a multi-level phenomenon
Since our mechanism-based explanation focuses on cognitive processes, we
must first appreciate that cognitive sciences are themselves plural and that the field is
interdisciplinary. This has been the case from the start, as one of its pioneers recalls:
“I argued that at least six disciplines were involved: psychology, linguistics,
neuroscience, computer science, anthropology and philosophy. I saw psychology,
linguistics and computer science as central, the other three as peripheral” (Miller,
2003, p. 143). Note that interdisciplinary endeavors may range from a mere
borrowing of concepts or tools to the establishment of a new inter-discipline with its
own discipline-like contents, structures and conventions (Klein, 1990). In our current
research, different types of interdisciplinarity will be involved simultaneously.
Somewhat simplifying, I mentioned earlier that there is a link between the
contribution of different disciplines and the different levels of a mechanism. This is
easier to see in non-organismic mechanisms, as they are not so complex nor flexible
in their organization. From cosmology via molecular physics to quantum physics, we
can distinguish different disciplines focusing on particular levels of mechanism. They
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -10
use their own tools and methods and formulate different theories and hypotheses.
Even though the differences between, for example, quantum mechanics, relativity
theory and classical mechanics are considerable, I believe that in the cognitive
sciences the characteristics of the different levels and the associated differences are
more wide-ranging. For connecting sociological and psychological observations,
brain images, neurochemical interactions and genetic factors, we need a wide variety
of conceptual and methodological tools, whereas in the physical sciences we can draw
on the assistance of mathematics.
Given this complexity and the interdisciplinary nature of our investigations,
decomposing action understanding into action recognition, intention understanding,
and narrative understanding will make our task more manageable. Our next step is to
localize these task components somewhere ‘in’ the individual or rather in that
person’s cognitive apparatus – foremost the brain. The interaction between cognitive
processes and their contexts may involve social scientific and humanistic
investigations. Applying our mechanism-based explanatory perspective to the
decomposed task of action understanding, we will once more approach it as a multi-
level phenomenon. Of course, we already did that when decomposing the task into
three components, but now we are heading for the brain and neural tissue. Indeed, we
must try to assign the components and operations of our phenomenon to concrete,
localizable, bodily or neural areas and activities. Generally, if we start from a
particular level, we can look both upwards and downwards. Looking downwards
“involves describing lower-level mechanisms for a higher-level phenomenon” where
these mechanisms are responsible for sub-tasks of the task or phenomenon (Craver,
2007, p. 257). Conversely, if we look upwards, we may be able to see our mechanism
interacting with other mechanisms that together realize a new phenomenon at this
higher level. For instance, our action understanding will, in interaction with
perception, motivation, and motor action, contribute to the agent’s response to another
person’s behavior. At that level, it can be considered a component alongside several
other components and operations. So in looking downwards we can treat action
understanding as an independent variable and detect changes in the associated
mechanisms. Or conversely, we can investigate the changes in action understanding
due to influences at a higher level, for example. Figure 2 at the end of this chapter
may illustrate these investigative approaches to the mechanism that we are
discovering in association with action understanding.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -11
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -12
Interdisciplinary collaborations are involved in the investigations of such a
mechanism, bringing different intervention and observation techniques with them.
Researchers can experimentally intervene in the components or operations at a
particular level and try to detect the consequences at another level -- upwards or
downwards. In the cognitive sciences -- including neuroscience -- such interventions
can be distinguished generally as interference, stimulation or activation experiments
(Craver, 2007). Stimulation involves, for example, the presentation of a stimulus to a
subject and detecting correlated activation at lower levels, down to single neurons.
Interference implies disturbing the normal mechanism, for instance by electro
stimulation of neurons, with subsequent detection of behavioral differences at a
higher level. Apart from such “vertically” directed interventions, researchers can try
to influence the mechanism “horizontally.” By varying the stimuli, researchers can
observe at various levels whether different mechanisms are activated, displaying
different properties in cognitive processing. Activation experiments can also be
combined with brain imaging techniques, which allow researchers to observe the
associated lower level activities of the posited mechanism, its components and
operations. The colorful images of brain activation patterns published in great
numbers are the results of such activation experiments, for instance. Furthermore,
there is also the valuable assistance of comparative work performed by ethologists,
developmental psychologists, computer simulation scientists, philosophers and again
social scientists and humanists. The latter disciplines can help to investigate whether
action understanding relies on different mechanisms in subjects from collectivist
versus individualist societies, or whether there is a difference in focus during
processing perceptual information between religious and non-religious subjects, for
instance. Clearly, limiting the number of disciplines to those that are most relevant to
the problem at hand is crucial to make the interdisciplinary research process
manageable – even though empirical evidence can lead to the need to include an
originally excluded discipline.7
7 Climate change models only recently include ocean dynamics, after marine scientists proved that these were involved in climate change, for example.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -13
3 Identify relevant disciplines to involve in the mechanism-based
approach
In general, deciding which disciplines are most relevant for any given research
project can be guided by three questions: “Does the discipline have a well-defined
perspective on the problem?”, “Has the discipline produced a body of research (i.e.
insights) on the problem of such significance that its published insights and
supporting evidence cannot be ignored?”, and “Has the discipline generated one or
more theories to explain the problem?” (Repko, 2008, pp. 169-170). Although action
understanding is instantiated by a multi-level mechanism, it is not surprising to find
that disciplines focused on very low levels—like quantum mechanics—have not
delivered useful insights to it. Fortunately, not all events at such low levels make
themselves felt at much higher levels in a relevant way. Quantum phenomena do
occur in every atom of the brain, but if they affect cognition or behavior they do so
only when they influence functioning of specific neural areas. A cosmological
phenomenon like sunspots may impact on human cognition only when it influences
earthly temperatures which have an impact on environmental conditions of humans,
which finally can affect human cognitive processes. However, it is rarely the case that
a single neuron seriously influences a cognitive process that involves many more
neurons, or that human cognition is directly and irrevocably influenced by
environmental temperature. Therefore and in accordance with the observation that
everywhere in the universe we find “local maxima of regularity and predictability”
(Wimsatt, 2007, p. 209), we can restrict our multi-level system investigations to the
nearest levels of our phenomenon.8 Even though interdisciplinarians must be critical
of the traditional division of labor among disciplines and keep an open eye to
contributions from unexpected disciplines, the fact that we can conceive of
connections between extremely divergent disciplines is never reason enough to
overlook differences in relevance and specificity.
Meanwhile, a first estimation of relevant disciplines for our research of action
8 It is also related to the intriguing phenomenon of ‘emergent properties’: properties that occur at a particular level and cannot be explained purely on the basis of our knowledge of lower level components and operations. We can now say that such emergent properties are likely to depend partly on the systemic interactions between components and operations at the particular level itself and even that higher level (“top-down”) contributions will often be involved, leaving little room for strictly reductionist explanations of higher level properties (Bechtel & Richardson, 1993).
understanding can be made. While our initial topic of action understanding implies
inclusion of the full range of the cognitive sciences, the social sciences, and the
humanities to account for all varieties of action understanding, our first delineation
and decomposition of it has made the research more manageable. Instead of one broad
phenomenon, we are now able to focus on three distinct components (action
recognition, intention understanding and narrative understanding), which will lead to
different questions, research methods and results, and theories.
Most straightforward is arguably the investigation of the first component,
action recognition, which is the capability of “parsing” or sequencing continuous
bodily behavior or movements into distinct actions. It is plausible both from a
developmental and an evolutionary perspective that this parsing capacity is present in
newborns and animals. Consequently, we may at first exclude the humanities and
even social sciences from investigations of this component of action understanding.
Again, it may turn out that later cognitive or speech developments affect the
mechanism that carries out action parsing, but our preliminary hypothesis is that
without these developments action parsing is still being performed. With the
exclusion of those sciences, there are still enough candidates for inclusion such as
neurophysiology, developmental psychology, biology, and information science.
For the second component of action understanding, intention understanding,
we may need to include social sciences and humanities in our investigations. Note
that we cannot straightaway exclude those sciences that assist to explain action
parsing. Parsing remains generally a precondition for understanding intention, and it
appears in some cases to contribute significantly to understanding the specific
intentions of actions. For frequently repeated actions, it has been argued that primates
derive the intentional structure of complex actions purely on statistical processing
(Byrne, 1999). However, it is implausible that such processing should suffice for all
instances of intention understanding. How about discovering the intentions of newly
observed actions, or of irregular and complex actions, involving tools? There is much
evidence of humans taking a so-called “intentional stance,” assigning to the observed
agent the possession of mental states such as beliefs, desires and reasons. This stance
is extremely useful for predicting future behavior of relatively autonomous agents
(Dennett, 1989). Experimental observations show that even young infants expect that
agents are aiming rationally at a particular goal and show surprise if their behavior
contradicts this expectation (Gergely, Nádasdy, Csibra, & Bíró, 1995). For explaining
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -14
these mental states, we need to draw in more scientific disciplines, because these
states involve other types of social information, often mediated by language, symbols,
and so on. Social sciences and humanities can systematically investigate the
interactions between such influences and action understanding. These influences are
even transmitted at lower levels of the mechanism, and are correlated with the
patterns of activity of so-called mirror neurons which have turned out to play an
important role in this domain. These mirror neurons were discovered some 20 years
ago and have surprising properties, because they respond both to perception and to
observation or imagination of actions. Interestingly, their activation depends partly on
prior experiences of the observers, even with socio-culturally specific information.
Understanding and imitating an action that is within their ‘vocabulary’ or actions is
consequently easier than if they observe it for the first time (Rizzolatti & Sinigaglia,
2008). For this component we can draw on philosophical analysis of intentional
action, on behavioral biology and psychology, on the cognitive sciences, and perhaps
on social scientific research that focuses on socio-cultural specific means and goals of
action.
In some instances of human action, goals, objects, and instruments are
perceptually visible, facilitating intention understanding. However, often the action is
not or incompletely visible, or the action is ambivalent or is temporally extended, or
the perceptually available information is sparse. Not surprisingly, therefore, our third
component of action understanding, narrative understanding, relies much more on
higher cognitive processes and the use of narrative structures. We employ language,
concepts, abstraction, temporal and causal relations and the like when developing
narrative structures. These are generally of an abstract nature and cannot be perceived
through the senses. With the help of such structures, “we comprehend other people’s
minds by creating a coherent narrative or story of their actions, organized around their
goals,” including the conditions, the agent’s plans and possible outcomes (Read &
Miller, 2005, p. 125). These coherent narratives are naturally based in part upon the
observable properties of agents and their actions, but also depend upon previous
expertise and knowledge of the observer, including relevant socio-cultural
information. While making use of the insights pertaining to intention understanding,
for this component we need also insights from the humanities and the social sciences
about the construction and use of narratives, and theories of meaning in speech and
behavior. Moreover, we may want to check for cognitive scientific explanations of
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -15
these phenomena as well, in order to explain why schizophrenics have difficulties
with delivering narratives, for instance.
In sum, simple explanations of action understanding are not to be expected.
Even with appropriate neural and cognitive functions, humans will face limitations in
their capability of understanding each other because of the variability due to socio-
cultural influences on individuals’ cognitive processes. Indeed, social cognitive
scientists or cognitive anthropologists argue that socio-cultural specific differences in
action understanding have “two distinct moments of birth, one public and
conventional and the other a subjective appropriation and integration of a
conventional form by a particular person. The links between public models and
personal knowledge are contingent relations” (Shore, 1996, p. 371). Such contingency
applies less strongly to action recognition, as such statistical and perceptual processes
will be more generally present in individuals from different cultures—even though for
socio-cultural actions we may need specific expertise to be able to parse complex
ritual actions. Given the number of disciplines listed above, interdisciplinary research
can benefit from prudently leaving out a discipline if it is not relevant to the specific
component or operation in which we are interested.
4 Conduct a literature search
Having exposed the scope and complexity of action understanding, it is apt to
conclude that it is not a research topic but rather a comprehensive field of
interdisciplinary research. However, with the help of the mechanism-based
explanatory approach, we are able to narrow the scope of our research adequately.
First, by decomposing the phenomenon we have helpfully delineated three component
tasks: action recognition, intention understanding and narrative understanding. Now,
we can try to limit ourselves to one of the three component tasks when we engage in
behavioral, ethological, developmental psychological, or other studies at the
phenomenon level. We can investigate the conditions and results of handling the task,
and subsequently explain what in fact the apparent task consists of. We can also
observe whether adults, infants and animals are doing comparably well, and whether
their results are correlated with their language capabilities. A more advanced subject
would be the relations between the component tasks. For instance, recognition of an
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -16
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -17
action is sometimes facilitated by understanding its intention. Such relations therefore
complicate the mechanism.
Second, after having observed the phenomenon “horizontally,” we may then
look at it as a multi-level mechanism and view it “vertically.” Earlier I noted that we
do not need to go very high or low in the investigations of associated levels. So no
quantum mechanics or cosmology, but neurology, perhaps neurophysiology, and
cognitive psychology should be our prime domains for the literature search.
Third, science in general and mechanism-based explanation in particular is
especially interested in behavior, changes, and modifications. If nothing ever happens
to a phenomenon, it is difficult to give a representation of its relevant mechanism.
When investigating such a mechanism researchers can induce changes by using the
inter-level experiments that make use of the interference, stimulation and activation
techniques referenced earlier. Such experiments are relevant to our research.
Literature that refers to exceptional cases or pathologies should be included with
caution because the flexibility of the brain hinders generalization from such cases to
normal cognitive processes. For instance, even with dysfunctional mirror neuron
systems, autistic patients may be able to reach some intention understanding.
Fourth, parallel to these studies and drawing heavily upon them, in the
cognitive sciences and elsewhere researchers increasingly use computational
simulations of a phenomenon. When focusing on a specific component or operation,
comparison with results from such simulations may be informative. For instance,
when a specific mechanism is simulated in a neural network program, it can also be
used for virtual –or in vitro– interference, stimulation, and activation experiments as
is carried out on living subjects. More extravagant still, when building and testing
humanoid robots, roboticists use the insights of action understanding research. Such
research may provide us with information regarding which cues facilitate humans to
understand robot actions. For instance, a form of eye contact and gaze following by a
humanoid robot seems to be a prerequisite for effective interaction by humans with
them.9
Fifth, the social sciences and humanities contribute foremost to the
“horizontal” investigation of action understanding itself. For instance, cross-cultural
research could deliver insights into differences in action understanding properties. It
9 See (Breazeal, 2004) for more on this.
is implausible that socio-cultural differences have a large impact on the cognitive
mechanism itself, even though there is evidence of the co-evolution of language and
cognition (Donald, 1991). It is plausible, nonetheless, that a singular complex and
dynamic organ as the brain can lead to highly divergent processing when it automates
and stores much of the expertise that it gathers under specific socio-cultural
conditions. Anyone who has observed a chess master, a musician or a hierograph
reader perform their exquisite skills may doubt whether they have the same brain and
use the same cognitive processes as we all do – but, yes, largely they do. Even if these
distinct capabilities are only the consequence of modulations of the mechanism by
such individual and external influences, the results are relevant enough to our inquiry.
Sixth, sometimes researchers have established a specific topic that appears to
be representative of the phenomenon under scrutiny. In the case of action
understanding, the study of imitation has turned out to be exemplary for it. Studying
imitation, we gain insights in “two relationships that are central to understanding
minds in general and human minds in particular: the relationship between perception
and action and the relationship between self and other” (Hurley & Chater, 2005, p.
48). Meanwhile, imitation has been studied in various animals, infants, adults, and
computer simulations. Such an example facilitates interdisciplinary research and
translation efforts enormously.
As it is only after a first acquaintance with the literature that you may be able
to decide about these matters, interdisciplinary research truly is “a decision-making
process that is heuristic, iterative, and reflexive” (Repko, 2008, p. 137).
5 Develop adequacy with respect to the relevant components,
operations and interactions of the mechanism
After identifying the relevant disciplines, we must develop adequacy in them.
Then we should be able to decide about their specific relevance, what kind of
knowledge and how much we need from each (Repko, 2008, pp. 189-190). In the case
of the component tasks of action understanding, the range of disciplines that are
involved differ, leading to narrow or wider—in the case of narrative understanding—
interdisciplinarity. This distinction reflects the methodological and conceptual
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -18
distance between the disciplines (Newell, 1998). Achieving adequacy in research that
involves wider interdisciplinarity and that leads to an integration of insights is
obviously more difficult.
Fortunately, it is possible to reach adequacy in the case of investigations of a
multi-level mechanism and its components and operations. This is due to the
aforementioned fact that in such a mechanism there are “local maxima of regularity
and predictability” (Wimsatt, 2007, p. 209) even though these maxima may
themselves be produced by complex mechanisms. The regular and predictable
properties of atoms, for instance, hide various underlying probabilistic quantum
mechanisms. Or, referring to the figure at the end of this text, investigations may
focus on the local maxima that are represented by particular components or operations
that are included in that figure without having to cover all the rest. As a consequence,
there are many theories that describe and explain quite specific properties of the
system, perhaps under specific conditions. Such ‘theoretical pluralism’ is common in
the life and cognitive sciences, granting each theory only a relative significance for its
domain (Beatty, 1997). Since we are interested in a particular phenomenon, action
understanding or one of its three component tasks, we are permitted or even obliged
to select those insights that contribute significantly to our understanding of that
phenomenon: its occurrence, the components and operations that instantiate it, the
conditions under which it occurs, modulating influences from other processes, and so
on. Clearly, these insights will be different in kind. Some will be based upon
observations of action understanding in humans, animals or even computer
simulations; others will refer to brain imaging results that suggest correlations
between specific components and operations involved in relevant cognitive processes,
for example.
Adequacy in our case must not imply presenting a complete mechanism-based
explanation that comprehensively predicts and explains action understanding under all
possible circumstances, as this would be extremely difficult. Instead, we have already
described how we can limit our research project to just a component or operation that
contributes to it. Having done so, we can subsequently aim first to develop a
mechanism sketch which explains how the phenomenon might be constituted. Such a
sketch leaves room for other sketches that offer different possible mechanisms for the
same phenomenon (Machamer, et al., 2000). Once we provide such a sketch—or
several sketches—starting from our preliminary definition, decomposition and
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -19
localization of action understanding, our investigations should allow us to gradually
fill in the details of our mechanism-based explanation. Adequacy then means that we
have included those insights that contribute specifically to the instantiation of our
research phenomenon, while leaving out others. Given the complexity and flexibility
of cognitive systems and the phenomena they produce, it is likely that future scientific
developments will have an impact on what insights need to be included. I will
illustrate these remarks pertaining to adequacy with an example of such a delineated
phenomenon.
We observed in the previous section that distinguishing “what” an action is, in
some cases delivers information on “why” it is performed as well. This suggests that
adequacy with respect to action recognition would also satisfy requirements for
adequate knowledge of disciplinary insights for intention understanding. Since it
turned out that animals and human adults and infants recognize the beginning and end
of an action by noting that body movements differ unexpectedly, changing in tempo
and direction (Baldwin, Baird, Saylor, & Clark, 2001), it seemed that adequacy would
be relatively easy to reach. After all, in this context achieving adequacy implies
gaining insight into a relatively simple perceptual mechanism, which performs
statistical processing. Moreover, the visual stimuli that appear to require processing
are only those that are associated with changes of tempo and direction. Consequently,
the number of components and operations that are involved in these cognitive
processes are limited. Consequently, the number of disciplines involved is limited,
and we are able to specify the insights that we need from them.
However, action recognition and intention understanding turned out not to be
two completely overlapping processes in many cases. Indeed, action recognition is not
always dependent upon perceptual processes alone, as it is often modulated by other
cognitive components. Conceptual knowledge of actions and specific task
requirements, like the command to focus attention on specific aspects of a movie,
enables subjects to recognize more reliably and faster the precise moments an action
begins and ends (Baldwin, et al., 2001; Hard, Lozano, & Tversky, 2006; Zacks,
Kumar, Abrams, & Mehta, 2009). Apparently, the action recognition mechanism can
be modulated by components and operations that subserve other cognitive processes.
Consequently, adequacy here requires additional knowledge of the mutual constraints
between the originally simple mechanism and the properties of such modulating
influences of other cognitive processes and the conditions under which these take
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -20
place.
This insight into the greater complexity of the action recognition mechanism
forces us to reconsider our striving for adequacy. At least it implies that we need to
refine or further specify the adequacy requirements with respect to disciplinary
insights. For instance, if we aim to keep the explanatory mechanism simple, we
probably need to refine more narrowly the action types that can be recognized solely
on the basis of this perceptual process of action recognition. Secondary to that, we
must investigate whether it is plausible that this process can function in isolation at
all. This seems to be the case in primate understanding and imitation of actions, where
allegedly supplementary understanding of aims, goals, and intentions of the agent is
generally not involved (Byrne & Russon, 1998). In humans this appears not a
plausible way to proceed because action recognition and intention understanding are
different yet more tightly connected phenomena in that case. Indeed, the former is
generally subserving the latter: “Such initial, ‘bottom-up’ parsing would provide
appropriate units on which to base the additional processing needed to achieve
ultimate understanding of the intentions at play. This type of low-level mechanism
thus seems likely to be a crucial prerequisite to infants’ developing understanding of
the intentions motivating others’ actions” (Baldwin, et al., 2001, p. 715). Adequacy
means in the case of human action recognition the investigation of whether observers
often rely upon previous socio-cultural knowledge or other cognitive processes to
recognize the borders of an action or between actions. If that is the case, the
explanatory mechanism for action recognition must be expanded, and our adequacy
requirements will be more severe, too.
Sometimes, an empirical finding suggests that adequacy is within reach. For
instance, the discovery of mirror neurons in the Macaque monkey motor system was
not just exciting, but appeared also to bring some relief to researchers of action
recognition and intention understanding. The peculiar activation of these neurons both
during action perception and during action performance suggested to many
researchers that action recognition and intention understanding could be adequately
explained at once by referring to these neurons. Even though these mirror neuron
systems are already more complex than the earlier proposed and purely perceptual
action parsing mechanism, they did appear at least to operate in isolation from higher
cognitive processes that involve speech. Indeed, they are still held to enable “that
modality of understanding which, prior to any form of conceptual and linguistic
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -21
mediation, gives substance to our experience of others” (Rizzolatti & Sinigaglia,
2008, p. 192). Still, the question remains: Was the promise fulfilled that adequacy
with respect to action recognition and intention understanding implied having insights
in the action parsing mechanism and mirror neuron systems only?
Unfortunately not, as it turned out that in human intention understanding, still
more is needed. Due to the complexity of our actions, humans cannot always
recognize action borders even with the additional help of these mirror neuron systems,
let alone the intentions of complex actions. Especially, intention understanding seems
to often rely on a “mentalizing” approach, which is the — silent and unconscious —
application of a “folk psychology” or “theory theory” that people use to explain or
understand “why” someone acts as he does. Such reasoning includes the use of
implicit psychological theories about (human) actions, goals, reasons, desires and the
like (Stich & Nichols, 1993). Even though this process does not yet involve explicit
and conscious verbalization, such an explanation of intention understanding does
involve many more cognitive processes than are produced by the action parsing
mechanism or the mirror neuron systems alone.
Not uncommonly, as soon as we include higher cognitive processes that
involve language or reasoning in the explanatory mechanism, adequacy will be
increasingly difficult to achieve. For instance, the “theory theory” account of intention
understanding has rival theories. One of these is a “simulation” theory which suggests
that subjects implicitly project themselves into the place of the agent, when observing
an action. The simulation theory claims to be supported by mirror neuron research
because these neurons allegedly enable such silent and immediate simulation (Gallese
& Goldman, 1998). The narrative approach proposes yet another and different take on
action recognition and intention understanding in humans. It refers to the fact that in
most cases, we ask agents themselves “why” they did “what” they did. It is then
“these second-person deliveries—the narratives narrated—that do the heavy lifting in
enabling us to understand and make sense of others with confidence” (Hutto, 2007, p.
21). Ricoeur—who taught us the distinction between the “what,” “why” and “who” of
action—explains that a narrative in fact establishes a sort of “plot” around an action,
which includes the character of the agent, the events experienced and those acted on.
Together, these allow us to establish the identity of agents and their actions alike,
even if such a narrative will never be complete or definitive (Ricoeur, 1992). Such
contributions of narrative to the explanation of our capacity of action understanding—
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -22
and of acting itself—has only quite recently gained the interest of philosophers and
scientists (cf. (Bayne & Pacherie, 2007; Gallagher & Hutto, 2008; Hutto, 2007;
Ricoeur, 1984; Zwaan, Taylor, & de Boer, in press, 2009).
Developing adequacy can thus take us in different directions. It can imply
revisiting the observations of the phenomenon itself, in order to find out whether we
can isolate a specific class of actions that can be recognized by a simple perceptual
mechanism alone. Or adequacy may require us to expand this mechanism with the
mirror neuron systems, still fencing out those cognitive processes that include speech.
Unfortunately, in humans this may still not yield adequacy as our narrative and
intention understanding depend mostly upon modulating factors that lie outside the
scope of these perceptual and mirror neurons systems. With respect to the disciplines
involved, adequacy requirements are enlarged because additional components and
operations need attention. On top of that, new disciplines—such as the social
sciences—need to be involved in order to reach adequacy. It is to be expected that in
our next step(s) we will feel the impact of this.
6 Analyze the phenomenon and evaluate each insight into it
As research is never simply the accumulation of factual knowledge, we now
need to analyze the problem from the perspective of each relevant discipline, and
evaluate each insight into it (Repko, 2008, p. 217). A mechanism-based explanation
allows us to respect the differences between disciplinary perspectives even though we
will assign disciplinary insights into a phenomenon only a limited role in the
comprehensive explanation. Disciplinary theories, methodologies and assumptions
may hold for the investigation of a component or operation of the phenomenon but
may have only limited relevance for the overall phenomenon.
Thinking of a phenomenon as the result of the interaction of a multi-level
organization of components and operations has the advantage that those components
and operations may be investigated separately. Of course, it is tempting to
disciplinary specialists to isolate their domain and to maintain that their domain of
study and the insights it delivers are sufficient to explain the phenomenon, leaving the
rest aside as irrelevant. In such a case, the assumption is that the phenomenon can be
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -23
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -24
explained with reference to a specific “module” which is relatively isolated and
independently responsible for all properties of the phenomenon.10 Even though I
argued earlier that there is relative autonomy of levels in a phenomenon, in organisms
there are many feed-back and feed-forward interactions at a particular level as well as
top-down and bottom-up interactions between levels that can influence such a
‘module’. We must remain on the alert for this possibility, indeed.
Analyzing and evaluating the disciplinary perspectives on the phenomenon of
action understanding within the mechanism-based explanatory approach is a
straightforward task. Starting from a mechanism sketch, mentioned in the previous
section, researchers will complete and describe the components, their operations and
indeed the interactions within the mechanism in ever more detail. This process
consists partly of deciding about the relevance of the contribution of different
components and operations, and the further arrangement of them. In our case, analysis
of action understanding was already part of step 1, where we defined and decomposed
it into the three component tasks. Clearly, achieving adequacy and this task of
analyzing, evaluating and arranging insights are intimately related research tasks that
need to be carried out repeatedly during the interdisciplinary research process.
The interrelated research tasks of analysis and evaluation as applied to the
present case of mirror neuron research can demonstrate that the tight connection
between the definition, decomposition and localization of a phenomenon, the
experimental set-up used in investigating it, and the subsequent results require much
care. An inadequate definition will hamper research as much as a bad experiment.
Take for instance the following conclusion drawn on the basis of a mirror neuron
activation experiment: “to ascribe an intention is to infer a forthcoming new goal, and
this is an operation that the motor system does automatically” (Iacoboni, Molnar-
Szakacs, Gallese, Buccino, & Mazziotta, 2005, p. 533). This sweeping conclusion
was drawn on the basis of an experiment in which humans were looking at images in
10 A famous example of such an assumption pertains to language, which has been claimed to rest in a specific mental organ, separate from the other mental organs and therefore also localizable in specific parts of the brain. These claims are highly implausible, if we take our earlier analysis of mechanism-based explanations for functions in complex and dynamic biological systems into account. And indeed, former language modularity proponent Chomsky, wrote in an influential review in 2002: “a neuroscientist might ask: What components of the human nervous system are recruited in the use of language in its broadest sense? Because any aspect of cognition appears to be, at least in principle, accessible to language, the broadest answer to this question is, probably, “most of it.”” (Hauser, Chomsky, & Fitch, 2002, p. 1570). In that review, the authors engage in a more modest approach, similar to the one I present here for action understanding.
which a hand would take a cup for drinking or for cleaning, and in which a breakfast
table was shown. Sure enough, the results were interesting as they suggested that our
brain automatically includes context information in its prediction of the likeliness of
the subsequent action with the cup. Nevertheless, the authors overstate the relevance
of their results by defining intention extremely narrowly, while applying a very broad
interpretation of the operation of the mirror neurons in this highly suggestive and
restrictive task. This exaggeration is partly due to the lack of a comprehensive,
interdisciplinary analysis of the phenomenon of intention understanding and
consequently the lack of a rigorous evaluation of the results. As a result, the
researchers suggest that these mirror neurons function as a module which in isolation
fulfills the difficult task of intention understanding. Indeed, one of these investigators
still maintains that these mirror neurons “embody the deepest way in which we stand
in relation to each other and understand each other” (Iacoboni, 2008). However, in
human relations we often experience our mutual deep involvement and attachment
through verbal meanings, which will not always activate mirror neuron systems. Such
meanings may heighten our sensitivity to some actions over others. In sum, although
we know that the evidence is convincing enough to include mirror neurons in the
action understanding mechanism, we still need other components to account for
intention understanding.
Such overstated claims are more common than one would expect. Research of
human social action, for example, can be characterized as being mostly divided along
two opposed theoretical positions, subscribing to the idea of ‘Plastic Man’ or
‘Autonomous Man’: “Whereas Plastic Man, being formed by adaptive response to the
interplay of nature and nurture, is only spuriously individual, his rival is to be self-
caused. (…) Where Plastic Man has his causes, Autonomous Man has his reasons”
(Hollis, 1977, p. 12). Such extreme positions are often due to overestimation of
disciplinary strengths and underestimating disciplinary limitations with respect to a
complex and dynamic, multi-level system. Fortunately, interdisciplinary exchanges
may force researchers to integrate their insights in a much more complex, but at the
same time more comprehensive explanation.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -25
B. Integrating insights
According to the model of the interdisciplinary research process that Repko presents
we would only now enter phase B, where the integration of insights takes place. Until
now, the focus was accordingly on “Drawing on disciplinary insights.” Even though I
have already looked ahead at the integration of these insights I will follow the
proposed research process and discuss the model’s next steps with respect to the
mechanism-based explanatory approach on offer here. Therefore, we need to identify
conflicts between insights and locate their sources (STEP 7) and then create common
ground between these insights (i.e., discovering one or more latent commonalities
between them) (STEP 8). Using this common ground we should be able to integrate
as many of the insights as possible (STEP 9). Eventually, this should bring us to a
more comprehensive understanding of human action and a way to it (STEP 10). The
mechanism-based explanation will prove a very useful instrument in these steps.
7 Identify conflicts between insights and locate their sources
Even though it is often the case that “[t]he possible sources of conflict
between insights are concepts, assumptions and theories” (Repko, 2008, p. 250), all
forms of interdisciplinary integration are not dependent upon resolving the conflicts
stemming from these building blocks of disciplinary perspectives on the phenomenon.
In the mechanism-based explanatory approach, conflicts of different types can occur.
One source of conflict stems from the initial phase of defining and decomposing our
phenomenon. In the previous section, we came across researchers who overstated
their conclusion, based upon an oversimplified definition of intention ascription.
A second source of conflict, related to such definition and decomposition
mistakes, is assuming that the underlying mechanisms of different component tasks
are completely separate. For instance, even though we distinguished the three
component tasks — action recognition, understanding of intention and narrative
understanding — we noted that it is likely that these components are intimately
related, both functionally and in their neural implementation.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -26
Different from such conceptual conflicts are those that arise from
misunderstandings of the internal arrangement of components and operations of the
mechanism. For instance, interactions between components or interactions have been
neglected or overlooked—as is the case when the component task of action
recognition would be forgotten as a prerequisite for intention understanding.
A fourth source of conflict is the underestimation of the role of specific
context features on the way an action is being cognitively processed. For instance,
only under favorable conditions –- without time pressure, for instance -- and with
adequate preparation are people able to suppress or rather overrule the power of
stereotypical modes of understanding other people (Kruglanski & Orehek, 2007).
Fifth, conflicts can arise from the failure to acknowledge and evaluate
correctly the alternative processing trajectories that may prevail in specific groups or
individuals when engaging in action understanding. For instance, it is still debated
whether or not autistic subjects, who have difficulties in spontaneous human action
understanding and often use specifically trained theories about human behavior, suffer
from disturbances in their mirror neuron systems, forcing them to rely on other
trajectories (Blakemore, Winston, & Frith, 2004).
A sixth source is a combination of the latter two sources of conflicts, as
external, socio-cultural information may have become entrenched in the explanatory
mechanism and cause observable and regular differences in action understanding
processing, as is the case with modulation of mirror neuron activity due to individual
experience with socio-cultural information (Keestra, 2008). In general, for
explanations of the variability of socio-cultural influences on individuals’ cognitive
processes, we need to draw on both cognitive and on social scientific insights (Shore,
1996).
This is not an exhaustive list of the possible sources and locations of conflicts
in mechanism-based explanations, but they are the most prominent. The list also
demonstrates that considering a phenomenon from a mechanism-based approach is
useful as a heuristic when reflecting on potential conceptual and theoretical failures
and conflicts. In our domain of action understanding, such conflicts have arisen
predominantly with respect to the component of intention understanding.
In particular, the rivalry between an explanation that is based upon tacit folk
psychological theorizing by the individual and an explanation that refers to the silent
simulation of the observed actor, has been quite fierce. Conflicts pertained to the
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -27
definition and decomposition of intention understanding, explanations of its
components’ tasks, the underlying mechanisms and their neural implementations in
the brain.
The opposition between these theorizing and simulation accounts was
modified somewhat as soon as tacit theorizing was no longer necessarily associated
with propositions, rules, psychological causal laws, and the like, because only then
were animals and infants equally able to understand intentions. So when connectionist
models were developed and computer tested with some success, theorizing accounts
seemed to gain the upper hand in the conflict (Stich & Nichols, 1993). The balance
shifted dramatically again, the moment the surprising evidence of mirror neurons
appeared. Simulation theorists immediately appreciated it as support for their position
and took it as evidence for a neural basis for their idea that we employ similar
structures to both actively engage in action and passively understand that action. This
was clearly stated in a seminal article co-authored by a neuroscientist and a simulation
theorist: “[t]hus M[irror]N[euron] activity seems to be nature’s way of getting the
observer into the same ‘mental shoes’ as the target—exactly what the conjectured
simulation heuristic aims to do” (Gallese & Goldman, 1998, pp. 497-498).
Extensive analysis of the history of this conflict between theorizing and
simulation accounts of intention understanding would show that the conflict
alternately received energy from theoretical arguments, from neuroscientific
evidence, from behavioral observations, and so on. As I note in the following
discussion, we can meanwhile observe that the two former opposing positions have
successfully been integrated into an overarching mechanism.
The mechanism-based approach allows us to analyze such a conflict and the
different sources from which it arises. Similarly, it can function as a heuristic device
that enables us to handle the conflict by employing its apparatus of phenomenon
analysis, components and operations, levels, and interactions of sorts. Most important
for interdisciplinary research is, of course, how the mechanism-based explanation can
be adapted in such a way that it can integrate those insights or properties that
appeared to be in conflict with each other, while both sides can present some evidence
in support. For that, we need to take the next step, where we are asked to create or
discover common ground.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -28
8 Create or discover common ground via a mechanism
Interdisciplinary investigations of human action understanding aim at the
integration of insights into that phenomenon. For such integration to succeed, it is
crucial to build upon a common ground that allows the integration of heterogeneous
materials. In our case this common ground should have the form of an explanatory
mechanism. Language processing is unlikely to provide such a ground, for instance,
as it cannot play a central role in explanations of action understanding in animals and
young infants. Nonetheless, given the strong interaction between components of any
comprehensive explanatory mechanism of action understanding, the explanatory
mechanism that we choose as common ground will be affected by the components
and operations that we will add to it. As noted above, language affects many other
components of the explanatory mechanism, including action recognition.
In fact, for most interdisciplinarians who use a mechanism-based approach
finding common ground for our interdisciplinary understanding involves the
identification of the most promising mechanism-based explanation among those that
have already been proposed in the literature. It is this mechanism-based explanation
as a whole that furnishes common ground. Note that this implies that generally such
common ground is already “composed of knowledge that is distributed among or is
common to disciplines” (Repko, 2008, p. 273).
After having identified a plausible explanatory mechanism as common ground,
interdisciplinarians need to demonstrate how other relevant components and
operations are related to it. What is implied in this endeavor is the rejection of claims
that a specific explanatory mechanism independently produces the phenomenon.
Embracing a mechanism as common ground can mean that it loses the exclusive
explanatory force it previously had. We have observed that mirror neuron systems
were taken to present the prime candidate for a mechanism-based explanation of
action understanding and imitation alike. Thanks to their functional properties, these
mirror neuron systems have been held responsible for enabling relationships between
oneself and another and between action and perception (Hurley & Chater, 2005).
Even though this makes these mirror neuron systems workable as common ground,
the associated mechanisms have meanwhile been embedded in a more comprehensive
mechanism, limiting their role accordingly. It is worth quoting a recent meta-analysis
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -29
that argues why still other systems or sub-mechanisms must be added to an
mechanism-based explanation of intention understanding with a central role for
mirror neurons:
First, an inconsistent or anomalous movement might be outside the perceiver's repertoire of familiar movements, so the mirror system cannot be of help. To resolve this, inferences of higher-level goals (e.g., “why did the actress fall?”) or other attributes (e.g., “was she depressed?”) seem to be needed, which are outside the scope of the mirror system. Second, when the perceiver reflects on a high-level intention of an action, this might necessarily engage the mentalizing system. It is possible that the mirror system is recruited for automatic lower-level goal interpretation (…) and the mentalizing system for reflection on the higher-level goals (from task goals to more general intentions). (Van Overwalle & Baetens, 2009, p. 580)
Nonetheless, various mechanism-based explanations of action understanding
have been presented in which mirror neuron systems function as common ground.
These systems offer then a common ground to different mechanism sketches or
models. Authors are sometimes even updating their own earlier model such as by
introducing a new model of action recognition learning by macaque mirror neurons which addresses data on auditory input, a model for opportunistic planning of sequential behavior, and studies of how to embed a macaque-like mirror system in a larger ape-like or human-like circuit to support ‘simple imitation’ and then ‘complex imitation.’ (Arbib & Bonaiuto, 2008, p. 45)
In this case, the previous model for imitation and understanding, with mirror
neuron systems as common ground, was expanded with components and operations
that process speech and symbol use. With that expansion, the authors are able to
explain the differences between human and monkey capabilities while leaving the
mechanism-based explanation of several commonalities largely intact.
What can be derived from this example is that, if possible, newly discovered
insights should be integrated into an existing explanatory mechanism. Obviously,
proposing a completely new mechanism is a much more demanding task. The various
adjustments of an existing explanatory mechanism in order to integrate insights into a
mechanism-based explanation are the subject of our next section.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -30
9 Integrate insights into a mechanism-based explanation
This chapter’s argument for a central role for mechanism-based explanation is
in accordance with the general observation that “At the heart of any interdisciplinary
integration lies an integrative device—for example, a metaphor, complex explanation,
or bridging concept—that brings together disciplinary insights” (Boix Mansilla,
Duraisingh, Wolfe, & Haynes, 2009, p. 344). I hope to have convincingly shown that
a mechanism-based explanation is a useful device for achieving interdisciplinary
integration. After having identified an appropriate explanatory mechanism as the
common ground for explaining our phenomenon, the remaining task is to build on this
when seeking integration of additional insights.
Disciplinary insights are in that case included as explanations of components
and operations of the mechanism, or as explanations of the interactions that take place
within the mechanism or between the mechanism and external factors. Given the
explanatory mechanism that we have identified as common ground, integrating
further insights will lead to refining or expanding it. Refining implies that we can
specify the mechanism of a sub-component or -operation or interaction of the
overarching explanatory mechanism. For instance, the action-perception interactions
that mirror neurons facilitate are affected both by action-specific experiences of the
subject and the task one is performing (Rizzolatti & Sinigaglia, 2008). Expanding the
mechanism involves an extension with a component, operation, or interaction which
has turned out to influence the mechanism in a relevant way. For instance,
investigations of socio-cultural influences on the cognitive processes that underlie
action understanding are relatively new and may lead to unexpected extensions: some
processes turn out not to be sensitive to these influences, others are strongly affected
by them (Han & Northoff, 2008).
Refining or expanding the explanatory mechanism with an additional
disciplinary insight requires a careful consideration of the place within the mechanism
where the addition could be localized. This is especially difficult in the case of a
complex and dynamic multi-level mechanism, where an addition is likely to have a
widespread impact on many components and operations. The figure below of the
explanatory mechanism of action understanding demonstrates the many choices
available for assigning a location to the additional insight. For instance, should we
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -31
consider a socio-cultural influence like religious belief as a ‘socio-cultural model’ that
functions as a specific ‘computation’ for narrative understanding, having only via that
route a modulatory influence on the perceptual mechanisms that produce action
recognition? Alternatively, we could investigate whether it is the imitation of
religious practices, implying mirror neuron system activations, that modulate
perceptual processes. So knowing that religion influences action understanding still
leaves undecided where and how this influence should be integrated into the
overarching explanatory mechanism. Similarly, we already mentioned that the action
understanding deficits in autistic patients may or may not be wide-ranging
consequences of disturbed mirror neuron systems (Blakemore, Winston, & Frith,
2004).
Integrating an additional insight by way of a refinement or expansion of the
explanatory mechanism that we chose as common ground therefore requires us to
specify one or more relations between previously unrelated components or processes.
The results of mirror neuron research have indeed forced researchers to reconsider
explanatory mechanisms for imitation, for action understanding, for empathy, for
language processing, and for action learning – to mention the most prominent
(Rizzolatti & Sinigaglia, 2008). Obviously, refining or expanding the mechanism-
based explanations of mirror neuron system activities will in that case require a
variety of techniques, depending on the specific interactions that the additional
component is involved in.
In this context, it may be useful to realize that Repko mentions four integrative
techniques that are commonly used for establishing common ground. These are: (1)
redefinition of concepts or assumptions to include or exclude phenomena; (2)
extension of concepts and assumptions or expansion of a theory to cover previously
uncovered phenomena, perhaps even beyond their original disciplinary domain; (3)
organization of previously unrelated concepts or assumptions into a relationship, and
finally (4) the transformation of opposing concepts or assumptions into variables of
an uncovered factor (cf. Repko, 2008, pp. 282-291, and the second edition of his
book). Obviously, common ground depends on establishing a relation between
previously unintegrated theories. According to the present approach it is advisable to
identify a given explanatory mechanism as common ground, and subsequently to
refine or expand this. Interestingly, the four techniques can in modified form also be
applied to these refinements or expansions as discussed below.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -32
For instance, the integrative techniques of redefinition and organization are
fairly common practice in the life sciences. They are often applied in combination as
can be learnt again from mirror neuron research. Among the many consequences of
the discovery of mirror neurons was the insight that the
rigid divide between perceptive, motor, and cognitive processes is to a great extent artificial; not only does perception appear to be embedded in the dynamics of action, becoming much more composite than used to be thought in the past, but the acting brain is also and above all a brain that understands. (Rizzolatti & Sinigaglia, 2008, p. xi)
Such rigid divides are also at stake when the integration of speech requires various
refinements and expansions of the mechanism-based explanation of mirror neuron
properties, which was proposed as common ground. These refinements and
expansions involve a re-organization of components and operations, comparable to
the third integrative technique of ‘organization of previously unrelated concepts or
assumptions into a relationship.’ Traditionally, language and action have been treated
as separate phenomena. Now that researchers realize that mirror neurons contribute to
both these cognitive processes, they are better able to explain the subtle and
sometimes disturbing interactions of language and action. This requires, however, a
redefinition (technique 1) of some basic assumptions concerning the ‘modular’
processes that were taken to underlie language and action.
Obviously, even when common ground has been established, further integration
of insights into the preferred mechanism-based explanation demands careful
application of various integrative techniques. Also, prudence and modesty are needed
with respect to the scope of the theories or explanations that require integration.
Researchers must realize that it is most unlikely that it is possible to localize cognitive
functions in particular neurons or mechanisms. Talk of neurons that ‘see’, or ‘feel’, or
‘infer’, should therefore be taken to mean that these components or operations play a
specific and decisive role in the comprehensive mechanism that accounts for that
function—no more, and no less (Keestra & Cowley, 2009).
To conclude, a general consequence of the integration of insights should be
acknowledged. Even if the focus is not on the assumptions or concepts of the
associated theories, integration will have an impact on the scope of these theories.
Integration of two theories has the consequence that their scope will become
expanded or contracted: expanded in the case when integration demonstrates that a
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -33
mechanism has also an impact on an additional component or operation, contracted
when the converse is true and a mechanism turns out to depend on another
mechanism for its operation, for instance. In cases where the integrated theory has
wide-ranging impact on the complex explanatory mechanism, both theory expansion
or contraction will obtain, though with respect to different components, operations or
interactions and the theories pertaining to these.
10 Present a mechanism-based explanation of human action
understanding and test it
Often, interdisciplinary understanding of a problem will be obtained at the final
stage of the research process. In a mechanism-based approach to explanation this is
somewhat different. As many—if not most—phenomena in the natural, life and
cognitive sciences do not appear contingently but reflect the behavior of a complex
mechanism, there are often already interdisciplinary mechanism-based explanations
available of components or operations of the eventual, overarching explanatory
mechanism. Remember the definition of a mechanism mentioned earlier: “[A]
mechanism is an organized system of component parts and component operations.
The mechanism’s components and their organization produce its behavior, thereby
instantiating a phenomenon” (Bechtel, 2005). When employing the mechanism-based
approach, researchers are aware of the fact that their specific research of such a
phenomenon should allow integration into such a mechanism as pertaining to a
component or operation or one of its interactions.
How we represent the explanatory mechanism is dependent upon its aim.
Complex and dynamic mechanisms, with their reciprocal constraints of components
and operations and their feed-back and feed-forward streams are difficult to represent
as a picture. Developing a computer program that includes such components and
operations is a more feasible method. However, if our interest is in the neural areas
that are involved in the mechanism, we may settle for a neuroanatomical map with
designated cortical areas and some broad processing streams. Or we might focus on a
finer grain of single neurons, like mirror neurons, and present the fine distribution of
those in a particular area and their connections. Often, researchers present their
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -34
findings in a relatively abstract ‘boxology’ where boxes and arrows refer to
components and operations. In that case, we can add labels to these that refer broadly
to their neural implementations in specific cortical areas. In the figure below, I have
presented most of the relevant information on action understanding contained in this
chapter in a mixed form of a boxology and the mechanism figure found in the
introduction. Clearly, it is far from complete and much more information could be
added to it or integrated with it. Specification of the components and operations
would require additional layers and theoretical descriptions of those but these are
beyond the scope of this chapter.
Finally, it is easy to see that such a visual representation of the mechanism can
assist further research in many ways, perhaps better than a verbally formulated theory
would do. Let me stress that as an integrative device mechanism-based explanation
does not exclude formulation of verbal and mathematical theories. Such theories often
focus on specific components or operations as we find them in the mechanism. As we
noted above, cognitive scientific experiments involve activation, interference or
stimulation of components or operations. The consequences of those can subsequently
be detected as changes in components or operations at other levels, or indeed in the
behavior of the subject. With the help of a mechanism sketch such as the one below
we can more specifically engage in the formulation of hypotheses and thinking of
potential interventions in it. We can consider horizontal interactions and their
potential effects, or we can reflect on inter-level investigations, using the interference
and stimulation techniques mentioned earlier. Such tests should lead to refinements,
adjustments or perhaps even the rejection of the proposed mechanism sketch.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -35
Cognition Phenomenon: task definition (level 0)
Behavioral/verbal response
Action understanding
Perception
Moti-vation
Action
InteractionsPerson C
Person A
Action perception
Person B
Proprioception
Task decomposition
Action recogni-tion
Intention understan-ding
Narrative understan-ding
Auditory (verbal) perception
Visual perception
Motion detec-tion
Motor simula-tion
Theorizing
Action configura-tion
Linguistic structures
Socio-cultural models
Causal knowledge
Action cognition
Prefrontal lobe areas
Mirror neuron areas
Synaptic signals
Speed dectec-tion
Compu- tations
Retinal image
Agent/ Figure detection
Cortical implementations (level -1)
Neuronal interactions (level -2)
Figure: a highly simplified and incomplete mechanism sketch of human action understanding. Many dynamical (feedforward & feedback, top-down & bottom-up) interactions are left out. Further components and operations, like attention, memory, and awareness could be added to the sketch as they modulate action understanding. Note that each component and interaction could in turn be investigated as a complex phenomenon on its own, requiring a mechanism-based explanation. Note as well, that components may play a role in other mechanisms and contribute to various phenomena simultaneously. This is the case with mirror neurons, for example. (© M. Keestra)
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -36
Concluding remarks
Action understanding is a complex cognitive capability performed by a complex
cognitive mechanism. As we learnt above, that mechanism is comprised of various
components and operations, that are themselves open to mechanism-based
explanations. In developing a mechanism-based explanation of action understanding,
we need therefore to integrate insights from various disciplines by attributing these to
components or operations or to specific interactions in which the mechanism
participates. Strikingly, although there are many non-linear processes involved and
notwithstanding the complexity of the overarching mechanism, the phenomenon of
action understanding displays relatively stable properties under specific conditions.
The method proposed in this chapter requires application of the heuristics of
definition, decomposition and localization. However, not all phenomena lend
themselves to this explanatory approach. If a phenomenon does not appear at all to
behave in an orderly fashion, or if it turns out to be impossible —even preliminarily—
to localize and identify components or operations, then we may have to look for
another approach. Fortunately for us, human action understanding is a phenomenon
suitable for this approach. By learning more about the complex and dynamic
mechanism that explains it, we can appreciate even more this crucial capability and
cope with its constraints and limitations. Indeed, rather than being satisfied with
explanations of its successes, it may be even more important in our global society to
acknowledge the fragility of our capability of understanding.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -37
Bibliography
Arbib, M. & Bonaiuto, J. (2008). From grasping to complex imitation: mirror systems on the path to language. Mind & Society, 7(1), 43-64.
Baldwin, D. A., Baird, J. A., Saylor, M. M., & Clark, M. A. (2001). Infants Parse Dynamic Action. Child Development, 72(3), 708-717.
Bargh, J. A., & Chartrand, T. L. (1999). The Unbearable Automaticity of Being. American Psychologist, 54(7), 462-479.
Bayne, T., & Pacherie, E. (2007). Narrators and comparators: the architecture of agentive self-awareness. Synthese, 159(3), 475-491.
Beatty, J. (1997). Why Do Biologists Argue like They Do? Philosophy of Science, 64, S432-S443.
Bechtel, W. (2005). The challenge of characterizing operations in the mechanisms underlying behavior. J Exp Anal Behav., 84(3), 313.
Bechtel, W., & Richardson, R. C. (1993). Discovering complexity: decomposition and localization as strategies in scientific research. Princeton, NJ: Princeton University Press.
Blakemore, S. J., Winston, J., & Frith, U. (2004). Social cognitive neuroscience: Where are we heading? Trends in Cognitive Sciences, 8(5), 216.
Boix Mansilla, V., Duraisingh, E. D., Wolfe, C. R., & Haynes, C. (2009). Targeted Assessment Rubric: An empirically grounded rubric for interdisciplinary writing. The Journal of Higher Education, 80(3), 334-353.
Breazeal, C. L. (2004). Designing sociable robots. Cambridge, MA: The MIT Press. Byrne, R. W. (1999). Imitation without intentionality. Using string parsing to copy the
organization of behaviour. Animal Cognition, 2(2), 63-72. Byrne, R. W., & Russon, A. E. (1998). Learning by imitation: A hierarchical
approach. Behavioral and Brain Sciences, 21(05), 667-684. Changeux, J. P., & Ricoeur, P. (2000). What makes us think? A neuroscientist and a
philosopher argue about ethics, human nature, and the brain (M. B. DeBevoise, Trans.). Princeton: Princeton University Press.
Craver, C. F. (2007). Explaining the Brain. Mechanisms and the Mosaic Unity of Neuroscience: Oxford U P.
Dennett, D. C. (1989). The intentional stance. Cambridge, MA: MIT press. Donald, M. (1991). Origins of the modern mind. Three stages in the evolution of
culture and cognition. Cambridge, MA: Harvard University Press. Dupre, J. (2001). In defence of classification. Studies in History and Philosophy of
Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences, 32(2), 203-219.
Gadamer, H. G. (2004). Truth and method (J. Weinsheimer & D. G. Marshall, Trans.). London: Continuum
Gallagher, S., & Hutto, D. (2008). Understanding others through primary interaction and narrative practice. In J. Zlatev, T. Racine, C. Sinha & E. Itkonen (Eds.), The shared mind: Perspectives on intersubjectivity (pp. 17–38). Amsterdam: John Benjamins.
Gallese, V., & Goldman, A. (1998). Mirror neurons and the simulation theory of mind-reading. Trends in Cognitive Sciences, 2(12), 493-501.
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -38
Gergely, G., Nádasdy, Z., Csibra, G., & Bíró, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56(2), 165-193.
Gergen, K. J., & Semin, G. R. (1990). Everyday understanding in science and daily life. In K. J. Gergen & G. R. Semin (Eds.), Everyday understanding: Social and scientific implications (pp. 1-18). London: SAGE.
Hacking, I., ed. (1983). Scientific Revolutions. Oxford, Oxford University Press. Hard, B. M., Lozano, S. C., & Tversky, B. (2006). Hierarchical Encoding of
Behavior: Translating Perception Into Action. Journal of Experimental Psychology: General, 135(4), 588-608.
Han, S. H., & Northoff, G. (2008). Culture-sensitive neural substrates of human cognition: a transcultural neuroimaging approach. Nature Reviews Neuroscience, 9(8), 646-654.
Hauser, M. D., Chomsky, N., & Fitch, W. T. (2002). The Faculty of Language: What
Is It, Who has it, and How Did It Evolve? Science, 298, 1569-1579. Hedstrom, P., & Swedberg, R. (1996). Social Mechanisms. Acta Sociologica, 39(3),
281-308. Hollis, M. (1977). Models of man: Philosophical thoughts on social action.
Cambridge: Cambridge University Press. Hurley, S., & Chater, N. (2005). Introduction: The Importance of Imitation. In S.
Hurley & N. Chater (Eds.), Perspectives on imitation: from neuroscience to social science (Vol. I & II). Cambridge, MA MIT Press.
Hutto, D. D. (2007). Folk psychological narratives: The sociocultural basis of understanding reasons. Cambridge, MA: The MIT Press.
Iacoboni, M. (2008). Mirroring people: The new science of how we connect with others. New York: Farrar Straus & Giroux.
Keestra, M. (2008). The Diverging Force of Imitation: Integrating Cognitive Science and Hermeneutics. Review of General Psychology, 12(2), 127-136.
Keestra, M., & Cowley, S. J. (2009). Foundationalism and neuroscience; silence and language. Language Sciences, 31(4), 531-552.
Klein, J. T. (1990). Interdisciplinarity history, theory, and practice. Detroit, MI: Wayne State University Press.
Kruglanski, A. W., & Orehek, E. (2007). Partitioning the Domain of Social Inference: Dual Mode and Systems Models and Their Alternatives. Annual Review of Psychology, 58(1), 291-316.
Lakoff, G., & Johnson, M. (1999). Philosophy in the flesh the embodied mind and its challenge to western thought. New York, N.Y.: Basic Books.
Machamer, P., Darden, L., & Craver, C.-F. (2000). Thinking about Mechanisms. Philosophy of Science, 67(1), 1-25.
Miller, G. A. (2003). The cognitive revolution: a historical perspective. Trends in Cognitive Sciences, 7(3), 141-144.
Newell, W. H. (1998). Professionalizing Interdisciplinarity. Literature Review and Research Agenda. In W. H. Newell (Ed.), Interdisciplinarity: Essays from the Literature. (pp. 529-563). New York: College Entrance Examination Board.
Poldrack, R. A., Sabb, F. W., Foerde, K., Tom, S. M., Asarnow, R. F., Bookheimer, S. Y., et al. (2005). The Neural Correlates of Motor Skill Automaticity. J. Neurosci., 25(22), 5356-5364.
Read, S. J., & Miller, L. C. (2005). Explanatory coherence and goal-based knowledge structures in making dispositional inferences. In B. F. Malle & S. D. Hodges
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -39
Keestra – Interdisciplinary research of action understanding – final draft, April 2010 -40
(Eds.), Other minds: How humans bridge the divide between self and others (pp. 124). New York: Guilford Press.
Repko, A. F. (2008). Interdisciplinary Research: Process and Theory. Los Angeles: Sage Publications.
Reyna, S. P. (2002). Connections - brain, mind, and culture in a social anthropology. London: Routledge.
Ricoeur, P. (1984). Time and narrative. (volume 1) (K. McLaughlin & D. Pellauer, Trans.). Chicago: University of Chicago Press.
Ricoeur, P. (1992). Oneself as another (K. Blamey, Trans.). Chicago: University of Chicago Press.
Ricoeur, P. (2008). From text to action: Essays in hermeneutics, II (K. Blamey & J. B. Thompson, Trans.). London: Continuum Press.
Rizzolatti, G., & Sinigaglia, C. (2008). Mirrors in the Brain: How Our Minds Share Actions and Emotions. Oxford University Press.
Sacks, O. (1995). An Anthropologist on Mars: Seven Paradoxical Tales. New York: Alfred A. Knopf.
Shore, B. (1996). Culture in Mind. Cognition, Culture, and the Problem of Meaning. Oxford: Oxford University Press.
Sperber, D. (1996). Explaining culture: a naturalistic approach. Oxford: Blackwell. Stich, S., & Nichols, S. (1993). Folk Psychology: Simulation or Tacit Theory?
Philosophical Issues, 3, 225-270. Szostak, R. (2004). Classifying science: phenomena, data, theory, method, practice.
Dordrecht: Springer. Tomasello, M. (1999). The Cultural Origins of Human Cognition. Cambridge, MA:
Harvard University Press. Van Overwalle, F., & Baetens, K. (2009). Understanding others' actions and goals by
mirror and mentalizing systems: A meta-analysis. NeuroImage, 48(3), 564-584.
Wimsatt, W. C. (2007). Re-Engineering Philosophy for Limited Beings. Piecewise approximations to reality: Harvard University Press.
Zacks, J. M., Kumar, S., Abrams, R. A., & Mehta, R. (2009). Using movement and intentions to understand human activity. Cognition, 112(2), 201-216.
Zwaan, R. A., Taylor, L. J., & de Boer, M. (in press, 2009). Motor resonance as a function of narrative time: Further tests of the linguistic focus hypothesis. Brain and Language, In Press, Corrected Proof.